From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.3) Gecko/20041020 Description of problem: In a multithreaded process, the resource usage of terminated threads is tracked in ->signal->utime/stime/etc. When the process dies, those counters are rolled into the parent process' cutime/cstime/etc. But as long as the process is alive, getrusage(2) is the only way to see them. They don't show anywhere in /proc, nor consequently in ps(1) or top(1). This is a problem with processes that spawn many threads which do work and then terminate. Version-Release number of selected component (if applicable): 2.6.9-1.675_EL How reproducible: Always Steps to Reproduce: 1. See attached rusage.c 2. cc rusage.c -lpthread 3. ./a.out & sleep 5 ; ps SH -p $! ; sleep 5 ; ps SH -p $! ; kill $! Actual Results: [1] 16894 PID TTY STAT TIME COMMAND 16894 pts/1 Sl 0:00 ./a.out 16894 pts/1 Rl 0:05 ./a.out PID TTY STAT TIME COMMAND 16894 pts/1 S 0:00 ./a.out [1]+ Terminated ./a.out Expected Results: Second "ps" should show 0:10 Additional info: See also my original description of the problem at http://www.uwsg.iu.edu/hypermail/linux/kernel/0409.0/0813.html. Please consider merging Albert Cahalan's and mine patches from 2.6.10-rc1: http://linux.bkbits.net:8080/linux-2.5/cset@1.1988.75.34 http://linux.bkbits.net:8080/linux-2.5/cset@1.1988.75.35
Created attachment 108630 [details] Test program
looks like good U1 material. adding to U1 blocker list.
Upstream patches by the bug reporter fixed this between 2.6.9 and 2.6.10: http://linus.bkbits.net:8080/linux-2.5/user=Lev_Makhlis/cset@4174ac12low7VnmD6QeTWOGO0A06nw?nav=!-|index.html|stats|!+|index.html|ChangeSet@-6M
This changeset is needed first: http://linus.bkbits.net:8080/linux-2.5/user=albert/cset@4174ac06gFZo2QwKwkbno1I5iNQrkA?nav=!-|index.html|stats|!+|index.html|ChangeSet
Created attachment 109951 [details] backport of those two upstream changesets to 2.6.9-5.EL # /proc/TGID/stat with multithreaded CPU time totals (#142989) %patch1702 -p1
This should not be on Roland plate anymore. He posted the patch to the list, back in January. I don't think there is anything else we can do to help.
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHSA-2005-514.html